Dr. Flaherty and a graduate student RA will conduct statistical analyses using Mx, Latent Transition Analysis, R, and Stata. Project personnel will run simulation studies in order to test the performance of the latent class model under conditions with varying degrees of conformance with assumptions of the model. Project personnel will perform secondary data analysis testing the latent class model (and longitudinal extensions) performance as well as to add to our knowledge of substance use, abuse and dependence. The second project period will continue to investigate the interaction between model performance, measurement quality and item scale. The past four years'work have involved interesting problems with model selection and multimodality in the likelihood function, as well as statistical power to detect the proper number of classes. Measurement and covariate issues will also be studied in longitudinal models, including the latent transition and associative latent transition models. Dr. Flaherty will travel to Richmond to meet with Dr. Neale to work on the grant once a year. They will also have conference calls. All research will be written up and submitted for publication. Software and scripts produced will be made freely available to the public.